What are objective and subjective fidelity criterias?

Digital image processing is essential in various fields, including photography, computer vision, and medical imaging. When evaluating the quality of processed images, it is essential to employ appropriate fidelity measures. Two commonly used approaches in assessing image fidelity are objective and subjective criteria.

Image fidelity criteria in digital image processing

Image fidelity measures are the criteria that determine the important aspect of an image and enable the measurement of the image quality. This allows us to calculate the difference between original and enhanced or reconstructed images.

The information required from the fidelity measures is application specific. The choice and significance of fidelity measures can vary depending on the specific application or use case in digital image processing.

Different applications may prioritize different aspects of image quality, and therefore, the selection of appropriate fidelity measures should align with the specific requirements and goals of the application.

As mentioned earlier, fidelity criteria can be divided into two classes in the image processing world.

• Objective fidelity criteria

• Subjective fidelity criteria

Objective fidelity criteria

The objective fidelity criteria or measures consists of equations that can be used to measure the amount of error in a processed or reconstructed image by comparing it directly with another image.

This is mostly used for images that can be created from a compressed data file or by using a restoration method.

Objective fidelity criteria are not necessarily correlated with our perception of image quality. However, they can be pretty useful in providing a comparison between different versions of the same image.

Some of the most common objective fidelity measures are:

• Root-mean-square error

• Root-mean-square signal-to-noise ratio

• Peak signal-to-noise ratio

Root mean square error

It is found by taking the square root of the error between the original and reconstructed image, divided by the total number of pixels in the image

Where N^2 represents the number of pixels in the image, r and c, represent the number of rows and columns, respectively. I stand for the original image, and I^ stands for the reconstructed image.

Root mean square signal-to-noise ratio

The reconstructed image is considered to be the signal, and the error is taken as the noise. The ratio gives off the mean error between the two.

r and c represent the number of rows and columns, respectively. I stand for the original image, and I^ stands for the reconstructed image.

Peak signal-to-noise ratio

The PSNR value gives the peak ratio between the two images by considering the reconstructed image as a signal and the error between the two images as noise.

Where L= the number of gray levels. For example, for 8-bits L=256

All these measures are mathematical equations that demonstrate the ratio of how an image has been changed and the difference between the original and the reconstructed image demonstrated by calculating an error between them both.

Subjective fidelity criteria

Subjective fidelity measures involve human observers who evaluate and compare images based on their visual perception and subjective judgment. This method is typically conducted through subjective experiments, where participants rate the quality of images.

The subjective fidelity measures provide a better and more accurate reflection of our visual perception than objective measures do. Moreover, they are not complex to calculate like the objective fidelity measures.

Some commonly used subjective fidelity measures include:

• Impairment tests: Subjects rate images in terms of how bad they are.

• Quality tests: Rating is done in terms of how well the images are.

• Comparison tests: Side-by-side evaluation is done of images.

Which subjective fidelity measure is most efficient

The comparison-type tests are considered to provide the most valuable and accurate results. These tests are easy to conduct, providing a relative measure between the two images. Thus, in turn, is the easiest test for most people to conduct.

Moreover, impairment and quality tests require an absolute measure with no point of reference. This makes it more challenging to determine the decision in an unbiased manner.